- Email: [email protected]
Cervical cancer in older women: A molecular analysis of human papillomavirus types, HLA types, and p53 mutations
Cervical cancer in older women: A molecular analysis of human papillomavirus types, HLA types, and p53 mutations Bobbie S. Gostout, MD,a Karl C. Podratz, MD, PhD,a Renee M. McGovern,b and David H. Persing, MDb Rochester, Minnesota OBJECTIVE: The purpose of this study was to evaluate cervical cancers in older women to determine whether they differed from tumors in younger women with respect to human papillomavirus types, frequencies of p53 mutations, and presence of a proposed high-risk HLA-DR2 haplotype. STUDY DESIGN: Cervical tissue was obtained from women undergoing surgical treatment of in situ or invasive carcinoma of the cervix. Viral and genomic deoxyribonucleic acid was extracted. The presence of human papillomavirus deoxyribonucleic acid was detected by polymerase chain reaction amplification. Viral subtypes were assigned by means of a combination of type-specific amplification and automated sequencing of the L1 region. The presence of p53 mutations was evaluated by direct sequencing of exons 5 through 9. The HLA-DR locus was screened for the presence of the high-risk DRB1*1501 allele by means of selective polymerase chain reaction amplification followed by agarose gel electrophoresis of HLA-DR2 types. RESULTS: Tumors from 39 women 62 to 85 years old were analyzed. Tumors from 104 younger women formed a reference group. Human papillomavirus 16 was found in 41% and 54% and human papillomavirus 18 was found in 10% and 12% of the tissue samples from older and younger women, respectively. The overall distributions of human papillomavirus types did not differ statistically between the groups. One of the 25 older patients tested had a p53 mutation. This tumor also had a positive test result for human papillomavirus 18. The DR*1501 allele was present in 33% of the older patients and 28% of the younger patients. The expected frequency of this allele in white Americans is 19.8%. The increased frequency of this allele among both older and younger women with cervical cancer was statistically significant (P < .05). CONCLUSIONS: We hypothesized that cervical cancer in older women might differ from that in younger women with respect to human papillomavirus types, natural host immunity, or the frequency of nonviral origins of the cancer. The findings show, however, that tumors from older women are extremely similar to those from younger women with respect to the human papillomavirus types present and the infrequent occurrence of p53 mutations. In addition we found that an HLA-DR allele that is associated with a risk of cervical cancer in younger women is also associated with risk in older women. These findings are most consistent with a model similar to that in younger women but with an unusually long latency for the transforming effect of the virus in some hosts. (Am J Obstet Gynecol 1998;179:56-61.)
Key words: Cervical cancer, older women, HLA type, human papillomavirus
The accepted paradigm of the progression of cervical neoplasia involves exposure to an oncogenic strain of human papillomavirus (HPV) during the second or third decade of life and gradual progression to invasive disease during a period between 10 and 15 years. The peak incidence of disease varies and is somewhat dependent on the use of screening and on access to treatment for preinvasive lesions in different populations.1 In the
From the Department of Obstetrics and Gynecologya and the Department of Laboratory Medicine and Pathology,b the Mayo Clinic and Mayo Foundation. Received for publication December 12, 1996; revised October 29, 1997; accepted November 10, 1997. Reprint requests: Bobble Gostout, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Copyright © 1998 by Mosby, Inc. 0022-5223/98 $5.00 6/1/87725
56
United States the peak incidence of cervical cancer occurs among women in their early 40s, consistent with the usual ages at exposure and the latency periods. Clearly, however, new disease continues to occur to an appreciable extent even into the seventh decade of life and beyond.2, 3 Because only a small fraction of new cases of cervical cancer occur in older women, little is known about the pathogenesis of the disease in this group. Generalizations in clinical and pathologic reviews of cervical cancer state that cervical cancer in postmenopausal women mimics the disease in premenopausal women, but on careful analysis it is clear that the published research on cervical cancer includes few women ≥62 years old. Because new disease seen during or after the seventh decade of life appears to challenge the paradigm for cervical cancer, we elected to analyze tumors from older women. We hypothesized that several factors could
Gostout et al 57
Volume 179, Number 1 Am J Obstet Gynecol
be important determinants of disease in older women: somatic mutations, variant HPV subtypes, and immunologic senescence. Regarding somatic mutations, a subset of cervical cancers is thought to arise through mutations, resulting in a non-virus-mediated pathway. Current estimates suggest that about 93% of tumors show evidence of an oncogenic HPV and about 7% show negative results for known HPV strains.4 Somatic mutations in the gene encoding p53 occur in some cancers and in many non-virus-related cervical cancer cell lines.5-8 Because the acquisition of somatic mutations is related to cumulative lifetime exposure to mutagenic events (including background rates of error in deoxyribonucleic acid [DNA] replication), we hypothesized that such a non-virus-mediated pathway may be more frequent among older patients with cervical cancer. Another possibility is that the various HPV subtypes produce different patterns of disease. The paradigm for cervical cancer reflects progression as we know it for disease related to HPV-16, the virus most commonly associated with cervical cancer. HPV-18–associated disease progression is thought to be similar to HPV-16–associated disease, although there are hints that the preinvasive stage may be more clinically silent and that the disease may be more aggressive.9, 10 At least 10 other HPV types are associated with cervical cancer, but these are less well characterized with respect to clinical behavior. A reasonable question is whether disease in the seventh decade of life and beyond is related to the same virus strains as disease in younger women or whether different virus strains with variant biologic traits play a role in late-onset disease. The effects of aging on the host response to HPV must also be considered. Immune responsiveness may be among the critical factors determining progression versus regression of early dysplastic lesions. In HPV-containing genital warts, the degree of CD4+ and CD8+ lymphocyte infiltration correlates with spontaneous regression of the warts.11 The risk for development of cervical cancer appears to be related to HLA type, further evidence of the role of immune function in this malignancy. Specifically, the 1501 allele at the HLA-DRB1 locus (DRB1*1501) confers an increased risk for dysplasia and cervical cancer, especially HPV-16–associated cervical cancer. The 1301 allele (DRB1*1301) in contrast is associated with a decreased risk for cervical cancer.12, 13 Aging is associated with a gradual decrease in immune responsiveness, manifested by increased susceptibility to infection, weaker responses to vaccination, and decreases in circulating mature T cells as a result of a decrease in CD8+ lymphocytes.14 Cervical cancer in older women is possibly related to aging-associated loss of the protective immunity in a chronic virus carrier.15 A discrepancy in the risk-related HLA alleles found in association with cer-
vical cancer between older and younger women with cervical cancer would support such an age-related decrease in immunocompetence as a factor. Material and methods All subjects were patients undergoing evaluation at Mayo Clinic. They were an older subset of patients from a group of volunteers participating in a study of cervical cancer approved by the Mayo Institutional Review Board. The age of 62 years was somewhat arbitrarily selected as the lower limit for the older sample on the basis of eligibility for some senior citizen benefits at that age. Of the 39 older women, 17 (43%) were from Minnesota and 22 were from 8 other states. Four patients (10%) were from Olmsted County, representing the local population rather than the referral practice at Mayo Clinic. A fragment of cervical tissue was collected from patients with surgically treated cancer (including carcinoma in situ). The tissue was placed directly into a liquid storage medium and immediately frozen at –40°C. DNA was extracted from the tissue sample with a commercially available extraction kit (Isoquick; ORCA Research, Inc., Bothell, Wash). One negative control (no DNA) tube was inserted for every 9 clinical samples. The resultant DNA was potentially a combination of viral DNA in the integrated and episomal state and human genomic DNA. The DNA and negative control tubes were stored in an amplicon-free freezer until they were thawed for analysis. Clinical information, including age, clinical staging and histopathologic diagnosis, sexual history, and previous gynecologic history, was abstracted from the records of the women. Polymerase chain reaction was used to detect HPV DNA in the samples. Primers designed to allow type-specific amplification of the E6 region of HPV-16 and HPV18 are as follows: • 5´CCACAGTTATGCACAGAGCTGCAAACAACTATACAT (HPV-16-E6-140-36D) • 5 ´ T T G T C C A G AT G T C T T T G C T T T T C T T C A G G A CACAGT (HPV-16-E6-465-36U) • 5´CTGTGCACGGAACTGAACACTTCACTGCAAGACAT (HPV-18-E6-153-35D) • 5 ´ C AT TA A G G T G T C TA A G T T T T T C T G C T G GATTCAACGG (HPV 18-E6-471-37U)
The primer name that follows each oligonucleotide sequence indicates the DNA target. This is followed by a number that describes the position of the 5´ base in the viral codon. The number of bases in the oligonucleotide followed by a U or D indicates the 5´ to 3´ direction and length of the oligonucleotide. These primers have been tested extensively with cell lines of known HPV type and more than 300 clinical samples. The specificity of the primers was verified by sequencing the amplified products. The L1 consensus primers MY09 and MY11 were used for detection of other HPV types.16
58 Gostout et al
July 1998 Am J Obstet Gynecol
Fig 1 Age at diagnosis of cervical carcinoma in 39 older patients (Note that the youngest group, 62-64 years, is a 3-year span, whereas the others are a 5-year span).
Polymerase chain reaction mixtures were prepared under hoods in a designated locked laboratory that is programmed for nightly ultraviolet decontamination. No polymerase chain reaction–amplified DNA is allowed in this preparatory laboratory. For HPV-16 and HPV-18 type-specific polymerase chain reaction 2 mmol/L magnesium chloride was used, and for the L1 consensus primers 1.5 mmol/L magnesium chloride was used. Templates were separated from primers by means of a wax layer (hot-start polymerase chain reaction) until the reaction cycles began with a 5-minute soak at 95°C. Positive and negative control preparations were inserted with each polymerase chain reaction setup (maximum of 35) in addition to the control tube for every 9 DNA extractions. No HPV DNA was detected in any of the negative control preparations. A Coy TempCycler II (Coy Corporation, Grass Lake, Mich) unit was used for 50 amplification cycles at 94°C, 55°C, and 72°C for 30, 30, and 60 seconds, respectively. This step was followed by a 5minute extension time at 72°C. Agarose gel electrophoresis, followed by ethidium bromide ultraviolet detection, was used to detect the amplification product of appropriate size. A chemiluminescent probe made from a full-length amplicon from HPV-16 E6 was applied to a Southern blot of appropriate agarose gels to detect any amplification product not visible on agarose gel electrophoresis (ECL Direct Nucleic AcidLabeling and Detection Systems; Amersham Life Science Inc, Arlington Heights, Ill). For HPV-18 a 184-bp internal oligonucleotide (bases 212 to 396) was used to detect low levels of HPV DNA with the same chemiluminescent system. Automated sequencing of the amplified L1 region was performed, allowing identification of the HPV types in these samples with computerized L1 DNA sequence comparisons. Direct sequencing of p53 exons 5 through 9 was accomplished by means of a previously described tech-
nique.17 The tissue-derived p53 sequence was then compared with the published sequence (and known polymorphisms) for the p53 gene. Suspected mutations were confirmed by performing a second amplification and sequencing procedure and by opposite-strand sequencing of the involved region of the gene. The HLA class II DRB1 allele was determined by DNA sequence, rather than by conventional serologic typing. This allows sequence-based assignment of allele subtypes, which may be important for predicting highly specific antigen reactions. Because HPV-16 and HPV-18 share many features but have important differences in their DNA sequences, subtle antigenic differences are possible; it is therefore important to understand the HLA alleles with the same fine level of sequence discrimination. After polymerase chain reaction amplification of the class II locus, hybridization was accomplished with labeled, sequencespecific oligonucleotide probes according to a previously published protocol.18 The HLA-DRB1 assignment was based on determination of allele-specific binding. Results Distribution of cases. Cervical samples were collected from 39 white women 62 to 85 years old (mean 70.5 years; Fig 1). A consecutive series of tumor samples from 104 younger women (28 to 61 years old) was analyzed for HPV and HLA type and was used as a reference group. Among the older women, 36 (92%) had squamous cell carcinoma, including 8 women with carcinoma in situ. One patient (3%) each had adenocarcinoma, adenosquamous carcinoma, and verrucous carcinoma. Sixty-nine (66%) of the younger women had squamous cell carcinoma, including 10 women (10%) with carcinoma in situ. Twenty-nine women (28%) had adenocarcinoma, including 2 (2%) with adenocarcinoma in situ. Three patients (3%) had adenosquamous carcinoma: 1 (1%) squamous cell carcinoma with sarcomatoid features, 1
Volume 179, Number 1 Am J Obstet Gynecol
(1%) neuroendocrine tumor resembling carcinoid, and 1 (1%) small-cell carcinoma. Detection of HPV. The distributions of HPV types in the specimens from the older women and from the reference group of younger women are shown in Table I. A representative gel with fluorescent signals for HPV detection is shown in Fig 2. The overall rate of HPV positive results by L1 consensus primer was 71% for the older group and 77% for the control group. HPV-16, the predominant virus subtype, was found in 16 (41%) of the older women and 57 (54%) of the younger women. HPV 18 was detected in 4 (10%) and 12 (12%) of the older and younger subjects, respectively. Two (5%) samples in the older group and 5 samples (5%) from the younger group had HPV types other than 16 or 18. Six samples in the older group (15%) and 6 samples from the younger group (6%) showed evidence for mixed infection. No virus was detected in 26% of the samples from older women and 18% of the samples from younger women. The group of tumors with no virus detected probably includes tumors that arose from a non–virus-mediated pathway but also samples in which no virus was detected because of sampling difficulties or previous treatment. The specimen for our analysis was removed after the usual pathologic specimens were obtained to ensure uncompromised pathologic interpretation, including frozen sections and permanent blocks. In some cases of microscopic tumor this approach meant that all tumor had been removed. In these cases a portion of cervix adjacent to the lesion was removed for analysis. The relative frequencies of the virus subtypes in the older women were not significantly different from those in the younger group (χ2 = 5.56, 5 degrees of freedom). When only the samples that contained an identifiable portion of the lesion were considered in the analysis, the relative frequencies of the viruses were changed minimally but the frequency of samples negative for HPV decreased to 14% in both older and younger groups. Analysis of the p53 gene. A total of 25 samples were suitable for p53 analysis (contained tumor cells). Samples excluded from p53 analysis included 11 cases of cervical intraepithelial neoplasia III, 1 case of microinvasive carcinoma in which we had tissue adjacent to the lesion but the actual lesion had been cut and held for pathologic interpretation, 1 case of microinvasive carcinoma diagnosed with conization elsewhere (confirmed by review of slides by Mayo Clinic pathologists) but with no residual lesion at the time of hysterectomy at Mayo Clinic, and 1 case of a complete response to neoadjuvant therapy. One mutation was detected in 1 of the 25 samples analyzed. The mutation was a cytosine-to-thymine transition resulting in an arginine-to-cysteine missense mutation in exon 8 of the p53 coding sequence. The patient was an 82-year-old woman with HPV-18–associated squamous cell carcinoma.
Gostout et al 59
Fig 2. Six patient samples were independently tested for HPV-16 DNA (lanes 1 through 6) and HPV-18 DNA (lanes 7 through 12). In lanes 13 through 18, the same 6 samples were tested with broadrange HPV primers. All amplification products are flanked by a DNA ladder that marks 100-bp increments. The first 2 samples show a positive signal for HPV-16 in lanes 1 and 2 (predicted band size 325 bp). The middle 2 samples show a positive signal for HPV-18 in lanes 9 and 10 but no signal for HPV-16 in lanes 3 and 4 (predicted band size 318 bp). HPV DNA is present in all 6 samples with the broad-range primers, indicating that the last 2 samples (lanes 17 and 18) have HPV of a type other than 16 or 18.
Determination of HLA-DR subtype. The DRB1*1501 allele was detected in 10 (36%) of the 28 HPV-associated tumors from older women and 13 (33%) of the 39 samples from the overall group of older women. This same allele was found in 29 (28%) of the 104 tumors from younger women (Table II). The expected frequency of this allele among white Americans is 9.9% (population frequency approximately 19.8%).19 Similar, although slightly higher, frequencies were found among Olmsted County residents tested as part of a population-based study (15.4% allele frequency, estimated population frequency 30.8%). The optimal reference population for Mayo Clinic surgical patients is difficult to discern because of the mixture of local and referred patients. Because only 10% of the older patients were Olmsted County residents, the national figure for whites may be a more accurate approximation of the true control population frequency of the DRB1*1501 allele. The frequency of DRB1*1501 among older women was significantly different from the expected frequency among white
60 Gostout et al
July 1998 Am J Obstet Gynecol
Table I. HPV types in cervical carcinoma* HPV-16 Age group (y) <62 ≥62
HPV-18
HPV-X
Mixed types
No HPV
Indeterminate†
No.
%
No.
%
No.
%
No.
%
No.
%
No.
%
Total
57 16
54 41
12 4
12 10
5 2‡
5 5
6 6
6 15
19 10
18 26
5 1
5 3
104 39
HPV-X,Type of HPV other than HPV-16 or HPV-18. *The types were similarly distributed among older and younger women (χ2 = 5.56, 5 degrees of freedom, P not significant). †Includes patients with discrepant results after repeated testing and patients with polymerase chain reaction products that
could not
be sequenced to verify DNA sequences consistent with HPV. ‡Includes 1 patient with HPV-33 and 1 with HPV-54.
Table II. DRB1*1501 allele in cervical cancer Present
Absent
Group
No.
%
No.
%
Total
<62 y ≥62 y ≥62 y, HPV-associated tumors ≥62 y, HPV-16–associated tumors
29 13 10 8
28 33* 36* 40*
75 26 18 12
72 67 64 60
104 39 28 20†
*In all 3 groups of women ≥62 years old the frequency of the DRB1*1501 allele was significantly increased (P < .05) compared with that among white Americans. †Includes patients in whom HPV-16 was isolated as a single virus type or as a component of a mixed infection.
Americans (χ2 = 4.5, P < .05) but not statistically different from the expected frequency in Olmsted County. The frequency of this allele among the younger women was similarly increased compared with the reference frequency among white Americans (χ2 = 4.27, P < .05) but not compared with that among Olmsted County residents. The frequency of this allele among younger women was not statistically different from that among older women with cervical cancer (χ2 = 0.38; 1 degree of freedom). Comment The scenario of a young woman seeking care for symptoms of a sexually transmitted disease is familiar to all practicing gynecologists. Thus the discovery of a sexually transmitted agent that promotes cervical dysplasia and cancer fits well with the usual demographics of patients with cervical cancer. The small set of older women in whom cervical cancer develops challenges the paradigm of virus exposure during the early years of coital activity followed by a 10- to 15-year progression from dysplasia to invasive cancer. A small subset of cervical cancers is believed to develop without the stimulus of HPV infection, and it would be easy to assume that cervical cancer in the older population represents just such a nonviral pathway to disease. We have shown, however, that the frequency and distribution of HPV types among an older population of patients with cervical cancer are not significantly different from those in a younger population treated in the same institution. Possible interpretations of this finding in-
clude (1) ongoing new exposure to sexually transmitted diseases into the seventh decade of life and (2) long latencies for HPV or its oncogenic effect in these older women. The sexual histories of one third of these women include long periods (40 to 60 years) of presumed mutual monogamy or abstinence late in marriage or after the death of a spouse. Because of potential inaccuracies in sexual histories, this information suggests but does not prove a minimal role for ongoing exposure to sexually transmitted diseases (data not shown). Also, the epidemiology of the more acute sexually transmitted diseases corroborates the notion that new exposures to sexually transmitted diseases in the seventh decade of life and beyond are uncommon. More likely, the spectrum of HPVinduced disease includes an extremely long latency in some hosts. In the stepwise progression toward malignant transformation, HPV infection is usually followed by viral DNA integration into the host genome. Additional, less wellcharacterized cellular changes follow, resulting in a fully transformed cancer cell. Possibly the accumulation of these events has a temporal probability that results in the usual 10 to 15 years of disease progression, with disease extremely early or late after exposure representing the normal tails of a bell-shaped probability distribution curve. The relationship of p53 mutations to cervical cancer initially seemed clear and complementary. Cervical cancer cell lines contained either p53 mutations or HPV
Volume 179, Number 1 Am J Obstet Gynecol
DNA, providing functional deletion of p53 cell-cycle regulation by one of two possible mechanisms in each cell line tested. However, the relationship of these factors in clinical samples turned out to be more complex. In cervical tumors p53 mutations are rare and show no clear relationship to the presence of HPV DNA in clinical samples. The sample of older women in our study demonstrates the same phenomenon. The only p53 mutation that was detected was found in an 82-year-old woman with HPV 18-positive squamous cell carcinoma. The relative contributions of these factors to disease pathogenesis in this individual are impossible to discern. There is a growing skepticism that host factors may play a role in an individual’s risk of acquiring cervical cancer after exposure to HPV. HLA typing may help find some individuals at risk. One demonstration of HLA-associated risk may involve the DRB1*1501 allele, which has been shown to be associated with HPV-16 cervical cancer among Hispanic American women from the southwestern United States.12 We found that the frequency of the DRB1*1501 haplotype in our older population was very similar to the frequency among younger women with cervical cancer treated at Mayo Clinic, and overall it was significantly higher than expected for white Americans. Thus at this time there is no evidence of a larger proportion of immunocompetent women in this older cohort. This finding challenges one of our hypotheses, that these older women may have had the benefit of immunologic control of a chronic viral carrier state and that their late-onset disease represents age-related loss of this biologic protection. The pathogenesis of cervical cancer in older women seems to be similar to that in younger women. It is reasonable to surmise that women who are exposed to oncogenic strains of HPV may face a certain risk of malignant transformation of infected cells for as long as they carry the virus. For the cohort of women in whom the virus is not efficiently eradicated after exposure, this risk represents an ever-present wild card that may present itself at any time in life. If the risk of cervical cancer continues for ≥40 years after exposure to HPV, recent epidemics of sexually transmitted diseases could have a considerable impact on the rates of cervical cancer far into the future. REFERENCES
1. Armstrong BK, Muñoz N, Bosch FX. Epidemiology of cancer of the cervix. In: Coppleson M, editor. Gynecologic oncology: fundamental principles-and clinical practice. 2nd ed. Edinburgh: Churchill Livingstone; 1992. p. 11-29.
Gostout et al 61
2. Kinney WK, Egorshin EV, Podratz KC. Wertheim hysterectomy in the geriatric population. Gynecol Oncol 1988;31:227-32. 3. Chapman GW Jr. Survival of advanced age females with cervical carcinoma. Gynecol Oncol 1992;46:287-91. 4. Bosch FX, Manos MM, Muñoz N, Sherman M, Jansen AM, Peto J, et al. Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. International Biological Study on Cervical Cancer (IBSCC) Study Group. J Natl Cancer Inst 1995;87:796-802. 5. Scheffner M, Munger K, Byrne JC, Howley PM. The state of the p53 and retinoblastoma genes in human cervical carcinoma cell lines. Proc Natl Acad Sci U S A 1991;88:5523-7. 6. Crook T, Wrede D, Vousden KH. p53 point mutation in HPV negative human cervical carcinoma cell lines. Oncogene 1991;6:873-5. 7. Fujita M, Inoue M, Tanizawa O, Iwamoto S, Enomoto T. Alterations of the p53 gene in human primary cervical carcinoma with and without human papillomavirus infection. Cancer Res 1992;52:5323-8. 8. Kurvinen K, Tervahauta A, Syrjanen S, Chang F, Syrjanen K. The state of the p53 gene in human papillomavirus (HPV)–positive and HPV-negative genital precancer lesions and carcinomas as determined by single-strand conformation polymorphism analysis and sequencing. Anticancer Res 1994;14:177-81. 9. Walker J, Bloss JD, Liao SY, Berman M, Bergen S, Wilczynski SP. Human papillomavirus-genotype as a prognostic indicator in carcinoma of the uterine cervix. Obstet Gynecol 1989;74: 781-5. 10. Rose BR, Thompson CH, Simpson JM, Jarrett CS, Elliott PM, Tattersall MH, et al. Human papillomavirus deoxyribonucleic acid as a prognostic indicator in early-stage cervical cancer: a possible role for type 18. Am J Obstet Gynecol 1995;173:1461-8. 11. Coleman N, Birley HD, Renton AM, Hanna NF, Ryait BK, Byrne M, et al. Immunological events in regressing genital warts. Am J Clin Pathol 1994;102:768-74. 12. Apple RJ, Erlich HA, Klitz W, Manos MM, Becker TM, Wheeler CM. HLA DR-DQ associations with cervical carcinoma show papillomavirus-type specificity. Nat Genet 1994;6:157-62. 13. Apple RJ, Becker TM, Wheeler CM, Erlich HA. Comparison of human leukocyte antigen DR-DQ disease associations found with cervical dysplasia and invasive cervical carcinoma. J Natl Cancer Inst 1995;87:427-36. 14. Lesourd BM, Meaume S. Cell mediated immunity changes in ageing, relative importance of cell subpopulation switches and of nutritional factors. Immunol Lett 1994;40:235-42. 15. Mandelblatt J. Squamous cell cancer of the cervix, immune senescence and HPV: is cervical cancer an age-related neoplasm? Adv Exp Med Biol 1993;330:13-26. 16. Ting Y, Manos MM. Detection and typing of genital human papillomaviruses. In: Innis M, Gelfand D, Sninsky J, White T, editors. PCR protocols: a guide to methods and applications. San Diego: Academic Press; 1990. p. 356-67. 17. Kovach JS, McGovern RM, Cassady JD, Swanson SK, Wold LE, Vogelstein B, et al. Direct sequencing from touch preparations of human carcinomas: analysis of p53 mutations in breast carcinomas. J Natl Cancer Inst 1991;83:1004-9. 18. Park MS, Tonai R. Phenotype frequencies of the class II (DR, DQ) DNA alleles by the patterns of sequence-specific primer mixtures (SSPM) in four different populations and the probable haplotypes between DRB1 allele and DQB1 allele. Clin Transpl 1992:475-500. 19. Tsuji K, Aizawa M, Sasazuki T, editors. HLA 1991: proceedings of the eleventh international histocompatibility workshop and conference. Oxford: Oxford University Press; 1992.
Key words: Cervical cancer, older women, HLA type, human papillomavirus
The accepted paradigm of the progression of cervical neoplasia involves exposure to an oncogenic strain of human papillomavirus (HPV) during the second or third decade of life and gradual progression to invasive disease during a period between 10 and 15 years. The peak incidence of disease varies and is somewhat dependent on the use of screening and on access to treatment for preinvasive lesions in different populations.1 In the
From the Department of Obstetrics and Gynecologya and the Department of Laboratory Medicine and Pathology,b the Mayo Clinic and Mayo Foundation. Received for publication December 12, 1996; revised October 29, 1997; accepted November 10, 1997. Reprint requests: Bobble Gostout, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Copyright © 1998 by Mosby, Inc. 0022-5223/98 $5.00 6/1/87725
56
United States the peak incidence of cervical cancer occurs among women in their early 40s, consistent with the usual ages at exposure and the latency periods. Clearly, however, new disease continues to occur to an appreciable extent even into the seventh decade of life and beyond.2, 3 Because only a small fraction of new cases of cervical cancer occur in older women, little is known about the pathogenesis of the disease in this group. Generalizations in clinical and pathologic reviews of cervical cancer state that cervical cancer in postmenopausal women mimics the disease in premenopausal women, but on careful analysis it is clear that the published research on cervical cancer includes few women ≥62 years old. Because new disease seen during or after the seventh decade of life appears to challenge the paradigm for cervical cancer, we elected to analyze tumors from older women. We hypothesized that several factors could
Gostout et al 57
Volume 179, Number 1 Am J Obstet Gynecol
be important determinants of disease in older women: somatic mutations, variant HPV subtypes, and immunologic senescence. Regarding somatic mutations, a subset of cervical cancers is thought to arise through mutations, resulting in a non-virus-mediated pathway. Current estimates suggest that about 93% of tumors show evidence of an oncogenic HPV and about 7% show negative results for known HPV strains.4 Somatic mutations in the gene encoding p53 occur in some cancers and in many non-virus-related cervical cancer cell lines.5-8 Because the acquisition of somatic mutations is related to cumulative lifetime exposure to mutagenic events (including background rates of error in deoxyribonucleic acid [DNA] replication), we hypothesized that such a non-virus-mediated pathway may be more frequent among older patients with cervical cancer. Another possibility is that the various HPV subtypes produce different patterns of disease. The paradigm for cervical cancer reflects progression as we know it for disease related to HPV-16, the virus most commonly associated with cervical cancer. HPV-18–associated disease progression is thought to be similar to HPV-16–associated disease, although there are hints that the preinvasive stage may be more clinically silent and that the disease may be more aggressive.9, 10 At least 10 other HPV types are associated with cervical cancer, but these are less well characterized with respect to clinical behavior. A reasonable question is whether disease in the seventh decade of life and beyond is related to the same virus strains as disease in younger women or whether different virus strains with variant biologic traits play a role in late-onset disease. The effects of aging on the host response to HPV must also be considered. Immune responsiveness may be among the critical factors determining progression versus regression of early dysplastic lesions. In HPV-containing genital warts, the degree of CD4+ and CD8+ lymphocyte infiltration correlates with spontaneous regression of the warts.11 The risk for development of cervical cancer appears to be related to HLA type, further evidence of the role of immune function in this malignancy. Specifically, the 1501 allele at the HLA-DRB1 locus (DRB1*1501) confers an increased risk for dysplasia and cervical cancer, especially HPV-16–associated cervical cancer. The 1301 allele (DRB1*1301) in contrast is associated with a decreased risk for cervical cancer.12, 13 Aging is associated with a gradual decrease in immune responsiveness, manifested by increased susceptibility to infection, weaker responses to vaccination, and decreases in circulating mature T cells as a result of a decrease in CD8+ lymphocytes.14 Cervical cancer in older women is possibly related to aging-associated loss of the protective immunity in a chronic virus carrier.15 A discrepancy in the risk-related HLA alleles found in association with cer-
vical cancer between older and younger women with cervical cancer would support such an age-related decrease in immunocompetence as a factor. Material and methods All subjects were patients undergoing evaluation at Mayo Clinic. They were an older subset of patients from a group of volunteers participating in a study of cervical cancer approved by the Mayo Institutional Review Board. The age of 62 years was somewhat arbitrarily selected as the lower limit for the older sample on the basis of eligibility for some senior citizen benefits at that age. Of the 39 older women, 17 (43%) were from Minnesota and 22 were from 8 other states. Four patients (10%) were from Olmsted County, representing the local population rather than the referral practice at Mayo Clinic. A fragment of cervical tissue was collected from patients with surgically treated cancer (including carcinoma in situ). The tissue was placed directly into a liquid storage medium and immediately frozen at –40°C. DNA was extracted from the tissue sample with a commercially available extraction kit (Isoquick; ORCA Research, Inc., Bothell, Wash). One negative control (no DNA) tube was inserted for every 9 clinical samples. The resultant DNA was potentially a combination of viral DNA in the integrated and episomal state and human genomic DNA. The DNA and negative control tubes were stored in an amplicon-free freezer until they were thawed for analysis. Clinical information, including age, clinical staging and histopathologic diagnosis, sexual history, and previous gynecologic history, was abstracted from the records of the women. Polymerase chain reaction was used to detect HPV DNA in the samples. Primers designed to allow type-specific amplification of the E6 region of HPV-16 and HPV18 are as follows: • 5´CCACAGTTATGCACAGAGCTGCAAACAACTATACAT (HPV-16-E6-140-36D) • 5 ´ T T G T C C A G AT G T C T T T G C T T T T C T T C A G G A CACAGT (HPV-16-E6-465-36U) • 5´CTGTGCACGGAACTGAACACTTCACTGCAAGACAT (HPV-18-E6-153-35D) • 5 ´ C AT TA A G G T G T C TA A G T T T T T C T G C T G GATTCAACGG (HPV 18-E6-471-37U)
The primer name that follows each oligonucleotide sequence indicates the DNA target. This is followed by a number that describes the position of the 5´ base in the viral codon. The number of bases in the oligonucleotide followed by a U or D indicates the 5´ to 3´ direction and length of the oligonucleotide. These primers have been tested extensively with cell lines of known HPV type and more than 300 clinical samples. The specificity of the primers was verified by sequencing the amplified products. The L1 consensus primers MY09 and MY11 were used for detection of other HPV types.16
58 Gostout et al
July 1998 Am J Obstet Gynecol
Fig 1 Age at diagnosis of cervical carcinoma in 39 older patients (Note that the youngest group, 62-64 years, is a 3-year span, whereas the others are a 5-year span).
Polymerase chain reaction mixtures were prepared under hoods in a designated locked laboratory that is programmed for nightly ultraviolet decontamination. No polymerase chain reaction–amplified DNA is allowed in this preparatory laboratory. For HPV-16 and HPV-18 type-specific polymerase chain reaction 2 mmol/L magnesium chloride was used, and for the L1 consensus primers 1.5 mmol/L magnesium chloride was used. Templates were separated from primers by means of a wax layer (hot-start polymerase chain reaction) until the reaction cycles began with a 5-minute soak at 95°C. Positive and negative control preparations were inserted with each polymerase chain reaction setup (maximum of 35) in addition to the control tube for every 9 DNA extractions. No HPV DNA was detected in any of the negative control preparations. A Coy TempCycler II (Coy Corporation, Grass Lake, Mich) unit was used for 50 amplification cycles at 94°C, 55°C, and 72°C for 30, 30, and 60 seconds, respectively. This step was followed by a 5minute extension time at 72°C. Agarose gel electrophoresis, followed by ethidium bromide ultraviolet detection, was used to detect the amplification product of appropriate size. A chemiluminescent probe made from a full-length amplicon from HPV-16 E6 was applied to a Southern blot of appropriate agarose gels to detect any amplification product not visible on agarose gel electrophoresis (ECL Direct Nucleic AcidLabeling and Detection Systems; Amersham Life Science Inc, Arlington Heights, Ill). For HPV-18 a 184-bp internal oligonucleotide (bases 212 to 396) was used to detect low levels of HPV DNA with the same chemiluminescent system. Automated sequencing of the amplified L1 region was performed, allowing identification of the HPV types in these samples with computerized L1 DNA sequence comparisons. Direct sequencing of p53 exons 5 through 9 was accomplished by means of a previously described tech-
nique.17 The tissue-derived p53 sequence was then compared with the published sequence (and known polymorphisms) for the p53 gene. Suspected mutations were confirmed by performing a second amplification and sequencing procedure and by opposite-strand sequencing of the involved region of the gene. The HLA class II DRB1 allele was determined by DNA sequence, rather than by conventional serologic typing. This allows sequence-based assignment of allele subtypes, which may be important for predicting highly specific antigen reactions. Because HPV-16 and HPV-18 share many features but have important differences in their DNA sequences, subtle antigenic differences are possible; it is therefore important to understand the HLA alleles with the same fine level of sequence discrimination. After polymerase chain reaction amplification of the class II locus, hybridization was accomplished with labeled, sequencespecific oligonucleotide probes according to a previously published protocol.18 The HLA-DRB1 assignment was based on determination of allele-specific binding. Results Distribution of cases. Cervical samples were collected from 39 white women 62 to 85 years old (mean 70.5 years; Fig 1). A consecutive series of tumor samples from 104 younger women (28 to 61 years old) was analyzed for HPV and HLA type and was used as a reference group. Among the older women, 36 (92%) had squamous cell carcinoma, including 8 women with carcinoma in situ. One patient (3%) each had adenocarcinoma, adenosquamous carcinoma, and verrucous carcinoma. Sixty-nine (66%) of the younger women had squamous cell carcinoma, including 10 women (10%) with carcinoma in situ. Twenty-nine women (28%) had adenocarcinoma, including 2 (2%) with adenocarcinoma in situ. Three patients (3%) had adenosquamous carcinoma: 1 (1%) squamous cell carcinoma with sarcomatoid features, 1
Volume 179, Number 1 Am J Obstet Gynecol
(1%) neuroendocrine tumor resembling carcinoid, and 1 (1%) small-cell carcinoma. Detection of HPV. The distributions of HPV types in the specimens from the older women and from the reference group of younger women are shown in Table I. A representative gel with fluorescent signals for HPV detection is shown in Fig 2. The overall rate of HPV positive results by L1 consensus primer was 71% for the older group and 77% for the control group. HPV-16, the predominant virus subtype, was found in 16 (41%) of the older women and 57 (54%) of the younger women. HPV 18 was detected in 4 (10%) and 12 (12%) of the older and younger subjects, respectively. Two (5%) samples in the older group and 5 samples (5%) from the younger group had HPV types other than 16 or 18. Six samples in the older group (15%) and 6 samples from the younger group (6%) showed evidence for mixed infection. No virus was detected in 26% of the samples from older women and 18% of the samples from younger women. The group of tumors with no virus detected probably includes tumors that arose from a non–virus-mediated pathway but also samples in which no virus was detected because of sampling difficulties or previous treatment. The specimen for our analysis was removed after the usual pathologic specimens were obtained to ensure uncompromised pathologic interpretation, including frozen sections and permanent blocks. In some cases of microscopic tumor this approach meant that all tumor had been removed. In these cases a portion of cervix adjacent to the lesion was removed for analysis. The relative frequencies of the virus subtypes in the older women were not significantly different from those in the younger group (χ2 = 5.56, 5 degrees of freedom). When only the samples that contained an identifiable portion of the lesion were considered in the analysis, the relative frequencies of the viruses were changed minimally but the frequency of samples negative for HPV decreased to 14% in both older and younger groups. Analysis of the p53 gene. A total of 25 samples were suitable for p53 analysis (contained tumor cells). Samples excluded from p53 analysis included 11 cases of cervical intraepithelial neoplasia III, 1 case of microinvasive carcinoma in which we had tissue adjacent to the lesion but the actual lesion had been cut and held for pathologic interpretation, 1 case of microinvasive carcinoma diagnosed with conization elsewhere (confirmed by review of slides by Mayo Clinic pathologists) but with no residual lesion at the time of hysterectomy at Mayo Clinic, and 1 case of a complete response to neoadjuvant therapy. One mutation was detected in 1 of the 25 samples analyzed. The mutation was a cytosine-to-thymine transition resulting in an arginine-to-cysteine missense mutation in exon 8 of the p53 coding sequence. The patient was an 82-year-old woman with HPV-18–associated squamous cell carcinoma.
Gostout et al 59
Fig 2. Six patient samples were independently tested for HPV-16 DNA (lanes 1 through 6) and HPV-18 DNA (lanes 7 through 12). In lanes 13 through 18, the same 6 samples were tested with broadrange HPV primers. All amplification products are flanked by a DNA ladder that marks 100-bp increments. The first 2 samples show a positive signal for HPV-16 in lanes 1 and 2 (predicted band size 325 bp). The middle 2 samples show a positive signal for HPV-18 in lanes 9 and 10 but no signal for HPV-16 in lanes 3 and 4 (predicted band size 318 bp). HPV DNA is present in all 6 samples with the broad-range primers, indicating that the last 2 samples (lanes 17 and 18) have HPV of a type other than 16 or 18.
Determination of HLA-DR subtype. The DRB1*1501 allele was detected in 10 (36%) of the 28 HPV-associated tumors from older women and 13 (33%) of the 39 samples from the overall group of older women. This same allele was found in 29 (28%) of the 104 tumors from younger women (Table II). The expected frequency of this allele among white Americans is 9.9% (population frequency approximately 19.8%).19 Similar, although slightly higher, frequencies were found among Olmsted County residents tested as part of a population-based study (15.4% allele frequency, estimated population frequency 30.8%). The optimal reference population for Mayo Clinic surgical patients is difficult to discern because of the mixture of local and referred patients. Because only 10% of the older patients were Olmsted County residents, the national figure for whites may be a more accurate approximation of the true control population frequency of the DRB1*1501 allele. The frequency of DRB1*1501 among older women was significantly different from the expected frequency among white
60 Gostout et al
July 1998 Am J Obstet Gynecol
Table I. HPV types in cervical carcinoma* HPV-16 Age group (y) <62 ≥62
HPV-18
HPV-X
Mixed types
No HPV
Indeterminate†
No.
%
No.
%
No.
%
No.
%
No.
%
No.
%
Total
57 16
54 41
12 4
12 10
5 2‡
5 5
6 6
6 15
19 10
18 26
5 1
5 3
104 39
HPV-X,Type of HPV other than HPV-16 or HPV-18. *The types were similarly distributed among older and younger women (χ2 = 5.56, 5 degrees of freedom, P not significant). †Includes patients with discrepant results after repeated testing and patients with polymerase chain reaction products that
could not
be sequenced to verify DNA sequences consistent with HPV. ‡Includes 1 patient with HPV-33 and 1 with HPV-54.
Table II. DRB1*1501 allele in cervical cancer Present
Absent
Group
No.
%
No.
%
Total
<62 y ≥62 y ≥62 y, HPV-associated tumors ≥62 y, HPV-16–associated tumors
29 13 10 8
28 33* 36* 40*
75 26 18 12
72 67 64 60
104 39 28 20†
*In all 3 groups of women ≥62 years old the frequency of the DRB1*1501 allele was significantly increased (P < .05) compared with that among white Americans. †Includes patients in whom HPV-16 was isolated as a single virus type or as a component of a mixed infection.
Americans (χ2 = 4.5, P < .05) but not statistically different from the expected frequency in Olmsted County. The frequency of this allele among the younger women was similarly increased compared with the reference frequency among white Americans (χ2 = 4.27, P < .05) but not compared with that among Olmsted County residents. The frequency of this allele among younger women was not statistically different from that among older women with cervical cancer (χ2 = 0.38; 1 degree of freedom). Comment The scenario of a young woman seeking care for symptoms of a sexually transmitted disease is familiar to all practicing gynecologists. Thus the discovery of a sexually transmitted agent that promotes cervical dysplasia and cancer fits well with the usual demographics of patients with cervical cancer. The small set of older women in whom cervical cancer develops challenges the paradigm of virus exposure during the early years of coital activity followed by a 10- to 15-year progression from dysplasia to invasive cancer. A small subset of cervical cancers is believed to develop without the stimulus of HPV infection, and it would be easy to assume that cervical cancer in the older population represents just such a nonviral pathway to disease. We have shown, however, that the frequency and distribution of HPV types among an older population of patients with cervical cancer are not significantly different from those in a younger population treated in the same institution. Possible interpretations of this finding in-
clude (1) ongoing new exposure to sexually transmitted diseases into the seventh decade of life and (2) long latencies for HPV or its oncogenic effect in these older women. The sexual histories of one third of these women include long periods (40 to 60 years) of presumed mutual monogamy or abstinence late in marriage or after the death of a spouse. Because of potential inaccuracies in sexual histories, this information suggests but does not prove a minimal role for ongoing exposure to sexually transmitted diseases (data not shown). Also, the epidemiology of the more acute sexually transmitted diseases corroborates the notion that new exposures to sexually transmitted diseases in the seventh decade of life and beyond are uncommon. More likely, the spectrum of HPVinduced disease includes an extremely long latency in some hosts. In the stepwise progression toward malignant transformation, HPV infection is usually followed by viral DNA integration into the host genome. Additional, less wellcharacterized cellular changes follow, resulting in a fully transformed cancer cell. Possibly the accumulation of these events has a temporal probability that results in the usual 10 to 15 years of disease progression, with disease extremely early or late after exposure representing the normal tails of a bell-shaped probability distribution curve. The relationship of p53 mutations to cervical cancer initially seemed clear and complementary. Cervical cancer cell lines contained either p53 mutations or HPV
Volume 179, Number 1 Am J Obstet Gynecol
DNA, providing functional deletion of p53 cell-cycle regulation by one of two possible mechanisms in each cell line tested. However, the relationship of these factors in clinical samples turned out to be more complex. In cervical tumors p53 mutations are rare and show no clear relationship to the presence of HPV DNA in clinical samples. The sample of older women in our study demonstrates the same phenomenon. The only p53 mutation that was detected was found in an 82-year-old woman with HPV 18-positive squamous cell carcinoma. The relative contributions of these factors to disease pathogenesis in this individual are impossible to discern. There is a growing skepticism that host factors may play a role in an individual’s risk of acquiring cervical cancer after exposure to HPV. HLA typing may help find some individuals at risk. One demonstration of HLA-associated risk may involve the DRB1*1501 allele, which has been shown to be associated with HPV-16 cervical cancer among Hispanic American women from the southwestern United States.12 We found that the frequency of the DRB1*1501 haplotype in our older population was very similar to the frequency among younger women with cervical cancer treated at Mayo Clinic, and overall it was significantly higher than expected for white Americans. Thus at this time there is no evidence of a larger proportion of immunocompetent women in this older cohort. This finding challenges one of our hypotheses, that these older women may have had the benefit of immunologic control of a chronic viral carrier state and that their late-onset disease represents age-related loss of this biologic protection. The pathogenesis of cervical cancer in older women seems to be similar to that in younger women. It is reasonable to surmise that women who are exposed to oncogenic strains of HPV may face a certain risk of malignant transformation of infected cells for as long as they carry the virus. For the cohort of women in whom the virus is not efficiently eradicated after exposure, this risk represents an ever-present wild card that may present itself at any time in life. If the risk of cervical cancer continues for ≥40 years after exposure to HPV, recent epidemics of sexually transmitted diseases could have a considerable impact on the rates of cervical cancer far into the future. REFERENCES
1. Armstrong BK, Muñoz N, Bosch FX. Epidemiology of cancer of the cervix. In: Coppleson M, editor. Gynecologic oncology: fundamental principles-and clinical practice. 2nd ed. Edinburgh: Churchill Livingstone; 1992. p. 11-29.
Gostout et al 61
2. Kinney WK, Egorshin EV, Podratz KC. Wertheim hysterectomy in the geriatric population. Gynecol Oncol 1988;31:227-32. 3. Chapman GW Jr. Survival of advanced age females with cervical carcinoma. Gynecol Oncol 1992;46:287-91. 4. Bosch FX, Manos MM, Muñoz N, Sherman M, Jansen AM, Peto J, et al. Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. International Biological Study on Cervical Cancer (IBSCC) Study Group. J Natl Cancer Inst 1995;87:796-802. 5. Scheffner M, Munger K, Byrne JC, Howley PM. The state of the p53 and retinoblastoma genes in human cervical carcinoma cell lines. Proc Natl Acad Sci U S A 1991;88:5523-7. 6. Crook T, Wrede D, Vousden KH. p53 point mutation in HPV negative human cervical carcinoma cell lines. Oncogene 1991;6:873-5. 7. Fujita M, Inoue M, Tanizawa O, Iwamoto S, Enomoto T. Alterations of the p53 gene in human primary cervical carcinoma with and without human papillomavirus infection. Cancer Res 1992;52:5323-8. 8. Kurvinen K, Tervahauta A, Syrjanen S, Chang F, Syrjanen K. The state of the p53 gene in human papillomavirus (HPV)–positive and HPV-negative genital precancer lesions and carcinomas as determined by single-strand conformation polymorphism analysis and sequencing. Anticancer Res 1994;14:177-81. 9. Walker J, Bloss JD, Liao SY, Berman M, Bergen S, Wilczynski SP. Human papillomavirus-genotype as a prognostic indicator in carcinoma of the uterine cervix. Obstet Gynecol 1989;74: 781-5. 10. Rose BR, Thompson CH, Simpson JM, Jarrett CS, Elliott PM, Tattersall MH, et al. Human papillomavirus deoxyribonucleic acid as a prognostic indicator in early-stage cervical cancer: a possible role for type 18. Am J Obstet Gynecol 1995;173:1461-8. 11. Coleman N, Birley HD, Renton AM, Hanna NF, Ryait BK, Byrne M, et al. Immunological events in regressing genital warts. Am J Clin Pathol 1994;102:768-74. 12. Apple RJ, Erlich HA, Klitz W, Manos MM, Becker TM, Wheeler CM. HLA DR-DQ associations with cervical carcinoma show papillomavirus-type specificity. Nat Genet 1994;6:157-62. 13. Apple RJ, Becker TM, Wheeler CM, Erlich HA. Comparison of human leukocyte antigen DR-DQ disease associations found with cervical dysplasia and invasive cervical carcinoma. J Natl Cancer Inst 1995;87:427-36. 14. Lesourd BM, Meaume S. Cell mediated immunity changes in ageing, relative importance of cell subpopulation switches and of nutritional factors. Immunol Lett 1994;40:235-42. 15. Mandelblatt J. Squamous cell cancer of the cervix, immune senescence and HPV: is cervical cancer an age-related neoplasm? Adv Exp Med Biol 1993;330:13-26. 16. Ting Y, Manos MM. Detection and typing of genital human papillomaviruses. In: Innis M, Gelfand D, Sninsky J, White T, editors. PCR protocols: a guide to methods and applications. San Diego: Academic Press; 1990. p. 356-67. 17. Kovach JS, McGovern RM, Cassady JD, Swanson SK, Wold LE, Vogelstein B, et al. Direct sequencing from touch preparations of human carcinomas: analysis of p53 mutations in breast carcinomas. J Natl Cancer Inst 1991;83:1004-9. 18. Park MS, Tonai R. Phenotype frequencies of the class II (DR, DQ) DNA alleles by the patterns of sequence-specific primer mixtures (SSPM) in four different populations and the probable haplotypes between DRB1 allele and DQB1 allele. Clin Transpl 1992:475-500. 19. Tsuji K, Aizawa M, Sasazuki T, editors. HLA 1991: proceedings of the eleventh international histocompatibility workshop and conference. Oxford: Oxford University Press; 1992.